Timestamps are in GMT/BST.
[0:00] <jay> i also wanted to add a seach feature (google search) to the chatlogs but i couldn't get it to work
[0:00] <jay> apparantly it doesn't work the way i think it does
[0:00] <jay> and google never seems to spider any urls i give it
[0:01] <Pascal> google is on a delay. Last time I submitted something it took 4 days
[0:01] <Pascal> that was about a month ago
[0:02] <jay> it was a month ago i submitted my personal site
[0:02] <Pascal> then they don't like you :)
[0:04] <jay> i knew that
[0:08] <Pascal> WTF?!?! rtMaxNodes: he number of unique nodes that can be contained in the routing table. Note that the node will try to keep an idle connection open to each of these, so don't set it to more than half the value of maxNodeConnections.
[0:08] <Pascal> then furthur down: rtMaxNodes = maxNodeConnections * 2
[0:09] <TheSeeker> lol
[0:10] <Pascal> rtMaxNodes: Too big or too small will result in inefficient or completely useless routing, or slow specialization;
[0:10] <Pascal> if thats true, then maybe we have a problem
[0:10] <TheSeeker> maybe that's why the thread limit isn't being adheared to...
[0:10] * e1ven (Kemrain@pool-68-163-146-123.bos.east.verizon.net) Quit ("Two")
[0:19] <TheSeeker> hmm, I'm seeing: %rtMaxNodes=100
[0:19] <TheSeeker> this could be because I haven't done a fresh install of the software for a long long time though, and have an ancient ini file...
[0:20] <salahx> coukld be
[0:20] <Pascal> default is 100, but it is overridden
[0:21] <Pascal> search for rtmaxnodes in node.java
[0:21] <TheSeeker> o.o
[0:21] <TheSeeker> *digs in the jar*
[0:22] <Pascal> it only obeys the configured (and default) value if maxNodeConnections <= 0
[0:22] <Pascal> in the jar? I'd like to see that. Only object code in the jar
[0:23] <TheSeeker> aw
[0:27] <Pascal> limit desc
[0:27] <Pascal> damn, wrong window
[0:28] <TheSeeker> if I use maxconnections at -1 won't that allow unlimited conenctions? o.o;;
[0:29] <Pascal> maxNodeConnections is forced to 0.4*maximumThreads unless maximumThreads = 0
[0:30] <TheSeeker> so to get a maxnodeconnections <= 0 you have to allow unlimited threads...
[0:32] <Pascal> thats what the docs say. havn't looked if its true or not
[0:34] <Pascal> wierd. "cat /proc/sys/fs/file-max" gives me 206412
[0:35] <Pascal> I thought 32k was the limit
[0:37] <salahx> maybe your initscripts increae it...
[0:37] <salahx> Its 47764 here
[0:39] <Pascal> pretty default FC2 install. Guess they increased it. ulimit still give me 1024 though. I'll have to increase that in my rc.local
[0:40] * jay (jcl@ool-18bf6dac.dyn.optonline.net) Quit (".")
[0:42] <Pascal> or I guess you are supposed to use profile.d now
[0:43] <salahx> 1024 is teh per-process limit
[0:43] <salahx> 206412 is teh GLOBAL SYSTEM limit
[0:44] <salahx> you;d change that in /etc/limits I think
[0:44] <Pascal> no such beast
[0:46] <salahx> its part of PAM
[0:46] <Pascal> lost me
[1:20] <Iakin> grr.. 95% of all my threads are waiting for trailerchunks..
[1:21] <Iakin> of those 98% will probably time out
[1:28] <TheSeeker> wheee!
[1:29] <TheSeeker> yay for overloaded nodes trying to send data to 200 people at once over a 12 KB connection!
[1:29] <salahx> mRI = Makes Routing Impossible :)
[1:31] <TheSeeker> heh
[1:34] * TheSeeker foolishly starts an unstable node on this POS computer
[2:07] <salahx> Hope you have a fire esgiusher handy...
[2:07] <TheSeeker> Transfers active (Transmitting/Receiving) 6 (0/6)
[2:07] <TheSeeker> but none of the connection icons show recieving data...
[2:13] * TheSeeker kills his RT and seednodes then re-starts
[2:15] <TheSeeker> now lets see if anyone remembers I was there a minute ago...
[2:15] <TheSeeker> yay, 1 incoming conenction
[2:16] <TheSeeker> and he's backed off, heh
[2:20] <TheSeeker> something very wrong with this:
[2:20] <TheSeeker> Connections open (Inbound/Outbound/Limit) 0 (0/0/55)
[2:20] <TheSeeker> Transfers active (Transmitting/Receiving) 1 (0/1)
[2:20] <TheSeeker> Data waiting to be transferred None
[2:20] <TheSeeker> how can I have an active transfer with no open connections?
[2:22] <TheSeeker> the one guy I know about is also 60167
[2:28] <TheSeeker> now I've got one icon showing that I'm transmitting, and the thing says I have 2 recieving and 0 transmitting x.x
[2:41] <vsalento> I have build jbigi/cpuid .so files and included them to freenet-ext.jar. However on startup I still got following message: INFO: Non-optimized native BigInteger library 'jbigi-linux-none.so' loaded from resource ... is this however better than non-jbigi impl?
[2:45] * taltipoa (~bla@195.25.40.253) Quit ("Leaving")
[2:53] * oierw (mathew@cpe-69-75-126-12.hawaii.rr.com) Quit (Remote closed the connection)
[2:53] * oierw (mathew@cpe-69-75-126-12.hawaii.rr.com) has joined #freenet
[3:02] <salahx> probably
[3:02] <salahx> You sohudl ahve a "libgi-linux-x86.so" though
[3:02] <vsalento> but there is no such thing
[3:03] <salahx> there shoudl ber, I think
[3:03] <salahx> where did you get it from ?
[3:03] <vsalento> and I would like to use one compiled for my gentooo if possible
[3:03] <vsalento> I got the contrib from the CVS and build it from there
[3:05] <salahx> i ahve'nt tried it yet
[3:05] <salahx> I ahve gentoo though
[3:07] <vsalento> I think I'll try to dynamically link to the gmp that I can emerge with gentoo
[3:08] <Pascal> http://dev.i2p.net/~jrandom/jbigi/
[3:15] <vsalento> pascal: thanks for the link but i need i586 version that doesn't exist even in there
[3:20] <Pascal> what is an i586?
[3:21] <salahx> a PEntium 1
[3:21] <vsalento> I have mini-itx computer that has those embedded VIA mainboards with some special cpu that doesn't require any fans
[3:22] <vsalento> I'm not 100% sure what it is but I need to compile stuff in gentoo as i586 for them to work
[3:24] <salahx> well libgi.c is merely "glue"
[3:26] <salahx> you might just be abkle to emerge gmp and restart Freenet; and that's that
[3:26] <salahx> I'm not exactly sure HOW it figured otu where the library is though
[3:29] <vsalento> there is tests available in java classes that can be used to see whether the native big integer calculations are better than java version. Those make it possible to see when it works.
[3:35] <salahx> i'm not even sure HOW to get the "glue"
[3:36] <salahx> so I ahveot remcomplile freenet-ext, or...?
[3:42] <vsalento> you need to get the Contrib module from the CVS... the glue is in there
[3:43] <salahx> thanks
[3:44] * tyesinclair (~tyesincla@170.252.64.1) has joined #freenet
[3:44] * tyesinclair (~tyesincla@170.252.64.1) has left #freenet
[3:52] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[3:53] <salahx> waht do I do with ti once I;ve compiled it ?
[3:55] <vsalento> just add the proper .so/.dll (depending on your os) to the freenet-ext.jar and restart the server
[3:55] <salahx> ok, so it goes into freenet-ext.jar trhen
[3:58] <vsalento> yes
[3:59] <salahx> how do I Get it in there?
[4:03] <salahx> hang on I think I got it now
[4:22] <vsalento> i updated the build.xml in freenet_ext directory to include both the cpuid/jbigi .so files into it
[4:23] <salahx> THere ANOTHER file ?
[4:24] <vsalento> another?
[4:24] <salahx> cpuid ?
[4:25] <vsalento> you need to compile the Contrib/jcpuid also to have working cpu detection code also
[4:25] <vsalento> there is simple build.sh to do it... it's quite simple
[4:41] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) Quit ()
[4:49] <vsalento> i get strange results in jbigi tests when comparing native run time to java run time. The results are:
[4:50] <vsalento> using jbigi-linux-none.so - modpow=1,50x (66%), double=1,85x (54%)
[4:50] <vsalento> using dynamically linked gmp (gentoo, i586) - modpow=5,78x (17%), double=1,79x (55%)
[4:51] <vsalento> if someone else has working native jbigi could you list how much faster it is for you (run the net.i2p.util.NativeBigInteger.main -method)
[4:56] <sanity> when we are ready to update the freenet-ext.jar in the snapshots directory let me know and i will do it
[5:01] <Pascal> sanity: why not update it nightly? The main build.xml now supports it.
[5:06] <toad_> <Pascal> toad: see what you think of http://bugs.gentoo.org/show_bug.cgi?id=45004 - uhm, when I run with NPTL enabled, even on 1.5.0b2, it hangs
[5:07] <toad_> <KenMan> that's quite possible. NPTL itself is still extremely highly alpha experimental code. Probably the widest used 'highly experimental alpha' stuff out there ! - no, that's linux 2.6 ;)
[5:09] <toad_> <salahx> Aren't timezone's wonderful?
[5:09] <toad_> <salahx> Stupid farmers....
[5:09] <toad_> farmers are smart enough not to need timezones
[5:10] <toad_> domestic parents and their brats are not
[5:11] <Pascal> toad: on what OS though?
[5:15] <Pascal> damn. When I moved my datastore from NetWare to Fedora all the timestamps were lost :(
[5:15] <oierw> yay hawaii. no daylight savings time. Of course, that's probably because they wouldn't understand how to change their clocks twice a year. (i'm sorry. hawaiians are, in general, not known to be the smartest of people but if I've offended anyone here, I'm somewhat sorry)
[5:16] <toad_> <Pascal> WTF?!?! rtMaxNodes: he number of unique nodes that can be contained in the routing table. Note that the node will try to keep an idle connection open to each of these, so don't set it to more than half the value of maxNodeConnections.
[5:16] <toad_> <Pascal> then furthur down: rtMaxNodes = maxNodeConnections * 2
[5:17] <toad_> the latter bit takes precedence
[5:17] <toad_> KenMan: just set TZ=Europe/London and be done with it
[5:18] <toad_> <Pascal> maxNodeConnections is forced to 0.4*maximumThreads unless maximumThreads = 0
[5:18] <toad_> not any more
[5:19] <Pascal> toad: then the docs need to be fixed
[5:20] <toad_> <Iakin> grr.. 95% of all my threads are waiting for trailerchunks..
[5:20] <toad_> <Iakin> of those 98% will probably time out
[5:20] <toad_> what makes you think that?
[5:20] <toad_> <TheSeeker> yay for overloaded nodes trying to send data to 200 people at once over a 12 KB connection! - Iakin has a rather bigger connection than 12kB
[5:21] <toad_> <salahx> mRI = Makes Routing Impossible :)
[5:21] <toad_> LOL
[5:21] <toad_> Transfers active (Transmitting/Receiving) 32 (24/8)
[5:21] <toad_> given a 20kB uplink that seems reasonable to me
[5:21] <toad_> maybe a bit high
[5:22] <toad_> but not insane
[5:24] <toad_> woah... Iakin, please add pentium 1 to the NBI targets?
[5:26] <toad_> ah, apparently there's already one
[5:26] <toad_> good
[5:26] <toad_> Pascal: yup
[5:27] <toad_> hmmm
[5:27] <toad_> how do i install a JVM on Mozilla?
[5:28] <Pascal> toad: what OS is NPTL broke for you on?
[5:29] <toad_> stock linux 2.6
[5:29] <Pascal> are we still keeping an idle connection open to every node in the routing table?
[5:30] <toad_> yup, trying to
[5:30] <toad_> what's wrong with that?
[5:30] <Pascal> I'm testing NPTL with -server now. So far looks fine
[5:30] <toad_> yeah, it'll take a day or two for it to hit it
[5:30] <Pascal> in node.java we set rtMaxNodes = maxNodeConnections * 2
[5:30] <toad_> or maybe not
[5:30] <toad_> ah
[5:30] <toad_> yeah
[5:30] <Pascal> so we are trying to keep open twice as many connections as max allowed
[5:30] <toad_> we keep a conn open to the top half of the routing table
[5:30] <toad_> roughly
[5:30] <Pascal> see any problem with that?
[5:30] <toad_> ordered by how useful they are
[5:31] <Pascal> so it won't thrash trying to keep them all open?
[5:32] <Pascal> pr think it is overloaded and refuse new because it is already at max?
[5:32] <Pascal> s/pr/or
[5:32] <toad_> no, it won't thrash.. not too much anyway
[5:32] <toad_> and no
[5:33] <Pascal> ok. Based on the docs I would have thought rtMaxNodes = maxNodeConnections / 2 woudl have been appropriate
[5:34] <Pascal> good news is my latest test node generated its own DSA group and so far no problems.
[5:34] <toad_> no, it's not anymore
[5:35] <toad_> if something is connected then it is in the RT
[5:35] <toad_> not the other way around
[5:35] <Pascal> ok cool. Another doc change :)
[5:37] * guido^pe (~unknown@dsl-213-023-232-066.arcor-ip.net) has joined #freenet
[5:45] <KenMan> damn... i've never seen my charts so well behaved... (partially due to the reset and lower number of users perhaps)
[5:49] <KenMan> http://mywebpages.comcast.net/jkcorson/FreeNet/60167.png and http://mywebpages.comcast.net/jkcorson/FreeNet/60167c.png
[5:50] <KenMan> i suppose if I give it long enough, things will come undone (?) ...
[5:55] <KenMan> anybody got more than 50 peers on unstable ???
[6:04] <Pascal> KenMan: yes
[6:09] <KenMan> Ah Hah! pu-pok authors CoE . I am certain of it !! ;)
[6:22] <toad_> hmm?
[6:23] <toad_> KenMan: sure, i've got 120
[6:23] <toad_> KenMan: why?
[6:23] <toad_> what makes you so sure?
[6:24] <toad_> woah, my ISO insert is still going...
[6:25] <toad_> i started it yesterday
[6:25] <toad_> somehow i managed not to restart my unstable node in the meantime
[6:29] <sanity> toad: your proposal seems reasonable (re: edge kludging)
[6:41] <toad_> hihihihihi
[6:42] <toad_> sanity: what about newbienes?
[6:43] <toad_> also do you think I should merge all the non-routing changes for 5085, and then do a 5086 when we're really sure about routing? the only reason to hold back the routing changes afaics is the edge kludging -> evil sort order issue
[6:43] <toad_> or should wait a few days and do the lot?
[6:51] * guido^pe (~unknown@dsl-213-023-232-066.arcor-ip.net) Quit (Read error: 104 (Connection reset by peer))
[6:51] * guido^pe (~unknown@dsl-082-082-158-159.arcor-ip.net) has joined #freenet
[6:53] <Pascal> wierd. For some reason File.renameTo() works between volumes on win2k, but not on Linux, but it still returns true.
[7:00] <sanity> toad: re: merges, yes
[7:00] <sanity> i think we should do it asap
[7:01] <sanity> toad: the issue with newbies is that they will get lots of requests?
[7:01] <sanity> ...because of low mris?
[7:02] <sanity> toad: one way to get more per-node samples is to have fewer nodes in the RT
[7:02] <toad_> no, the issue with newbie nodes
[7:02] <KenMan> well, it IS one way to get more samples...
[7:02] <toad_> how long do we keep them newbie?
[7:02] <toad_> arguably we should keep them newbie until they are no longer influenced by the startup bias, right?
[7:03] <toad_> but that could be a LONG time
[7:03] <toad_> especially if for some reason they happen to have a long MRI
[7:03] <sanity> remind me how we treat newbie nodes differently?
[7:03] * tyesinclair (~tyesincla@170.252.64.1) has joined #freenet
[7:03] <KenMan> they get discovery probes for one thing
[7:03] * tyesinclair (~tyesincla@170.252.64.1) has left #freenet
[7:04] <toad_> sanity: we don't drop them from the RT
[7:04] <toad_> also they can announce
[7:04] <sanity> toad: ah yes
[7:04] <toad_> even though they're in the RT
[7:04] <toad_> probably some other bits and pieces
[7:05] <toad_> atm we don't use newbieness for deciding whether to do discovery probes
[7:05] <toad_> but we could
[7:05] <KenMan> node management is the key one
[7:05] <toad_> oh yeah
[7:05] <toad_> we don't accept more than N% of the RT as newbie
[7:05] <toad_> if we have more we reject incoming conns
[7:05] <sanity> toad: how long do nodes stay nwebie right now?
[7:06] <toad_> sanity: if connected, 5 minutes or 20 routingHits
[7:06] <sanity> wasn't the point of newbie nodes to let them announce and then kick them off (unless they were proving useful)?
[7:06] <toad_> problem is that doesn't count QRs
[7:06] <toad_> so I changed it to 5 minutes or 200 totalHits
[7:06] <toad_> well, I think we need to keep them until they're competing on a level playing field
[7:07] <toad_> and in the case of really newbie nodes, that's especially important
[7:07] <sanity> in the case of pDNF, that could be a long time
[7:07] <toad_> hopefully they'll have announced and gotten useful in the meantime
[7:07] <toad_> just announcing doesn't mean they're useful
[7:09] <toad_> well, I think perhaps we want a 200 hit threshold for global smoothing (or whatever the mechanism is going to be called)
[7:09] <toad_> and kick them off around 200 hits
[7:09] * edt (~knoppix@dsl-130-81.aei.ca) Quit ("Client exiting")
[7:11] <toad_> that'll only be a few hours at worst, hopefully
[7:12] <sanity> ok
[7:12] <toad_> that would be 200 total hits however
[7:12] <toad_> also, there is the issue i raised on devl - what if the node has a high MRI?
[7:12] <toad_> do we REALLY want to keep it in the RT until 200 * veryVeryVeryLargeMRI ?
[7:13] <sanity> well what is the current criteria for dropping nodes from the RT apart from newbieness?
[7:14] <toad_> if it's a newbie, it doesn't get dropped
[7:14] <toad_> we drop according to the sort order
[7:14] <toad_> the sort order is:
[7:14] <toad_> 1. is it useless because of old version? if so, first head on the block
[7:14] <toad_> 2. is it a newbie? if so, last
[7:15] <toad_> 3. is it connected? if so, that's best. otherwise, go by the count of consecutive failed connections to it
[7:15] <toad_> 4. mri * typicalEstimate
[7:15] <toad_> that's it
[7:15] <toad_> :)
[7:16] <sanity> what was wrong with just dropping which ever node we had not sent a request to for the longest period of time?
[7:16] <toad_> well, we have to have the sort order for deciding which nodes to connect to
[7:16] <toad_> so we use it for everything
[7:17] <sanity> so a node's actual effectiveness is only considered last when deciding which node to drop?
[7:17] <toad_> hmm?
[7:17] <sanity> typicalEstimate is only considered in 4)
[7:17] <toad_> yup
[7:17] <toad_> if the node isn't connectible, there's no point considering its effectiveness
[7:18] <toad_> and there's no way to take that into account in the estimate, nor should there be
[7:18] <toad_> so that justifies 3.
[7:18] <sanity> just dropping which ever node we had not sent a request to for the longest period of time would account for all of those factors
[7:18] <toad_> 1. is justified because a node might become a wrong version after being a good version for some time
[7:18] <toad_> sanity: maybe it would, but then we'd still have to decide who to open connections to
[7:19] <sanity> toad: you already have a way to do that
[7:19] <toad_> IIRC we used to drop the LRU idle connection..
[7:19] <toad_> with some caveats
[7:20] <sanity> the current scheme seems much more complicated without any obvious benefits over the simpler scheme, and could well have some serious issues which are obscured by its complexity....
[7:20] <toad_> which would still be a problem because we would need to KEEP the current scheme
[7:20] <toad_> if only for deciding which nodes to connect to - AND which nodes to drop from the routing table
[7:20] <sanity> just for establishing connections, not dropping nodes
[7:20] <toad_> no, for dropping from the RT, as well as for opening conns
[7:21] <toad_> because the RT is a superset of the open conns
[7:21] <toad_> and IMHO it should remain so
[7:21] <sanity> i thought we connect to all nodes in the RT?
[7:21] <toad_> nope
[7:21] <toad_> we connect to half the nodes in the RT
[7:21] <sanity> why not?
[7:22] <toad_> why shouldn't the RT be larger than the nodes we actually can connect to?
[7:22] <toad_> it avoids various nasty problems we've been snagged on
[7:22] <toad_> e.g. node enters RT, node sucks, node eventually is thrown out, node re-enters RT, ...
[7:22] <sanity> hmmmmmmmmm
[7:22] <sanity> fuck, i have to go meet a friend for coffee
[7:22] <toad_> seeya :)
[7:23] <sanity> my closing thought is that we should try to simplify this, it seems unnecessarily complicated
[7:23] <sanity> we can follow up on the ML
[7:23] <toad_> what appears simpler often doesn't behave simpler
[7:23] <toad_> then you have to add complexity to fix it :(
[7:24] <toad_> sure, post something
[7:24] <toad_> seeya
[7:59] <vsalento> in inboundRequests diagnostic page the succeed column is 0 for all... why?
[8:02] <Pascal> toad: when my test node eventually locks up, what do I have to do to see if it is your NPTL bug?
[8:07] <toad_> Pascal: well, if it's got all threads blocked on one or two locks, which never become unlocked, despite not being locked by any threads, over multiple stack dumps, it's probably the same bug
[8:09] <toad_> bbl
[8:10] <Pascal> but how do I tell?
[8:15] <toad_> umm, take some stack dumps
[8:15] <Pascal> on linux, how?
[8:16] <TheSeeker> 'morning
[8:17] <Pascal> morning
[8:19] <toad_> kill -QUIT `cat freenet.pid`
[8:19] <toad_> INSERTING_BLOCKS (144/192):[1/6] queued: 29 running: 19
[8:19] <toad_> ummm
[8:19] <toad_> does that mean it's still on the first segment?!
[8:20] <toad_> or that it's doing segment 1, which may not be the first segment in the random insert order?
[8:20] <Pascal> so the -QUIT tells it not to actually kill the process but just get a stack trace?
[8:20] <toad_> Pascal: yup
[8:20] <Pascal> cool. thanks
[8:27] <Pascal> figured out the File.renameTo() problem. It does correctly return false on Linux, but the text of the error message was split so I couldn't find it. DistributionServlet is still broke, but now I'm too tired to fix it. Will do tomorrow.
[8:28] <Pascal> sleep
[8:39] * robilad (~topic@mpiat2304.ag2.mpi-sb.mpg.de) has joined #freenet
[8:40] <TheSeeker> I've got one node with an estimator graph that looks kinda like __|^|___ :P isn't that backwads of what it should look like?
[8:41] <TheSeeker> it's running 60167 btw
[8:45] <sanity> TheSeeker: its probably an artifact of the algorithm rather than an "anti"-specialization
[8:47] * Gunnar (~gunnar@217-13-11-42.dd.nextgentel.com) has joined #freenet
[8:48] * TheSeeker looks at the room topic
[8:48] <TheSeeker> should that be changed to 60167, or just wait 'till 168 :P
[8:50] <TheSeeker> well, I've been up for 6 hours... any useful stats I can pull off my node before shutting down to have resources avaliable to do more important stuff on this POS?
[8:50] <TheSeeker> or is 6 hours too short to have any meaningful stats?
[8:50] <sanity> TheSeeker: http://127.0.0.1:8888/servlet/nodestatus/psuccess_data.txt
[8:51] <sanity> TheSeeker: values will probably be around 0.01 on average - what are they for you?
[8:55] * greycat (~wooledg@192.35.79.70) has joined #freenet
[8:57] <TheSeeker> .07 to .011
[8:57] <TheSeeker> err
[8:57] <TheSeeker> .007 to .011
[8:58] <TheSeeker> when running fora day or so I usually get .04 to .06 on my main PC (6 gig datastore though, that may be why)
[9:13] * Ash-Fox (Hal-9000@aak18.neoplus.adsl.tpnet.pl) Quit (Read error: 110 (Connection timed out))
[9:16] * Ash-Fox (Hal-9000@aak131.neoplus.adsl.tpnet.pl) has joined #FREENET
[9:34] <toad_> hi sanity
[9:34] <sanity> hi toad
[9:35] <hirvox> hrm. how do I remove a directory from CVS?
[9:36] <toad_> you don't afaics
[9:40] <hirvox> there's now Contrib/jcpuid/lib/freenet/support/cpuinformation and Contrib/jcpuid/lib/freenet/support/CPUInformation (the latter is the correct one) in Contrib
[9:43] <toad_> sanity: i see no message from you on devl on simplifying routing
[9:44] <sanity> simplifying routing?
[9:44] <toad_> sanity: btw, imho one of the best ways to simplify things would be to have a fixed small block size :)
[9:44] <toad_> sanity: what we were talking about earlier
[9:45] <sanity> you mean simplifying RT pruining?
[9:46] <toad_> something like that
[9:47] <toad_> you were talking about connecting to all nodes in the RT too
[9:47] <sanity> well, the core issue was why we aren't just dropping the least recently contacted node
[9:51] <toad_> you mean least recently requested
[9:51] <toad_> in any case the difficulty with that was that we need another criterion for dropping nodes (the above is for disconnecting)
[9:51] <sanity> yes
[9:52] <sanity> well, how about least recently connected?
[9:53] <toad_> hmm?
[9:54] <toad_> that's just a proxy to our who-do-we-connect-to algorithm
[9:54] <sanity> so whichever has been disconnected for the longest amount of time gets droppped
[9:54] <sanity> well, the existing approach is probably ok for that
[9:55] <sanity> i mean, what is the underlying problem right now that is causing us to reevaluate this?
[9:55] <toad_> ummm
[9:55] <toad_> i don't remember :)
[9:55] <toad_> something to do with edge kludging
[9:56] <toad_> edge kludging was going to lead to nodes staying in the rt for aaaaaaages
[9:56] <toad_> err
[9:56] <sanity> oh yeah
[9:56] <toad_> removal of edge kludging was
[9:57] <sanity> well, ok - so solution: disconnect from least recently routed to node, and remove from routing table based on current scheme
[9:57] <toad_> why is that better than using the current scheme throughout?
[9:57] <toad_> i'm not convinced that such a hybrid is simpler than what we do now
[9:58] <sanity> well, you were the one who felt that what we do now wouldn't be able to handle the removal of edge kludging...
[9:58] <sanity> perhaps we should remove edge kludging, and see what happens...
[9:59] <toad_> do you agree that eventually we will have to use smaller maximum filesize, in the interests of a) make routing simpler, b) defeat traffic analysis, c) give routing more to go on and therefore more accuracy...?
[9:59] <sanity> yes
[9:59] <toad_> and d) more requests per hour => nodes cease to be newbies faster
[9:59] <toad_> do you agree that for primarily b), we will need a fixed small filesize eventually?
[9:59] <sanity> for example, in the current freenet-like software I am currently working on, all data is 100k
[10:00] <sanity> the main problem with it is that for small files such as webpages, it can be wasteful (min 100k per page)
[10:00] <toad_> yep
[10:00] <toad_> i was thinking 4k, 8k or 16k
[10:00] <toad_> fix it at one of the above
[10:00] <sanity> wow, that is small
[10:00] <toad_> advantage with larger is less cpu usage
[10:01] <toad_> advantage with smaller is more efficient for small files
[10:01] <toad_> and better against traffic analysis
[10:01] <sanity> smaller may not play too nice with TCP (it takes a while to get going)
[10:01] <TheSeeker> multiple of sector size is a good thing (tm)
[10:01] <toad_> sanity: huh?
[10:01] <sanity> what is the per-block bandwidth overhead?
[10:02] <toad_> you mean for the request chain?
[10:02] <sanity> tcp starts slow then speeds up, with small blocks it probably wouldn't get a chance to reach full speed before the block has been sent
[10:02] <toad_> it can be made pretty low
[10:02] <toad_> ah
[10:02] <toad_> yeah
[10:02] <toad_> well that's not really relevant because we reuse the conn
[10:02] <sanity> perhaps
[10:02] <toad_> 4kB might be a single packet in TCP if it wasn't for the fscking ethernet MTU :|
[10:03] <sanity> 16k seems really small to me, i would say 100k - what is the non-data bandwidth overhead per request?
[10:04] <toad_> well it's just that making them all 16k would be very wasteful for web pages
[10:04] <toad_> and frost messages
[10:04] <toad_> and a few other things
[10:04] <toad_> overhead... hrrm
[10:04] <toad_> a few hundred bytes for storables
[10:04] <toad_> a DataRequest
[10:04] <toad_> which I think is 120 bytes now
[10:04] <toad_> an Accepted
[10:04] <toad_> which is ~ 20 bytes
[10:05] <toad_> lots of QueryRestarted's, potentially
[10:05] <toad_> which are ~ 20 bytes each
[10:05] <toad_> a DataReply
[10:05] <toad_> which includes the storables
[10:05] <toad_> DataReply
[10:05] <toad_> TrailerMuxID=45ed
[10:05] <toad_> Storable.Initial-digest=fde57bd8e54ee6d5f3c9da9d4d917df8722845a0
[10:05] <toad_> Storable.Part-size=4000
[10:05] <toad_> Storable.Document-header=10667784e14f82d6aa02f1a0de928dc4b35d7379eaa209c678bf2d1
[10:05] <toad_> db604baa7
[10:05] <toad_> Storable.Symmetric-cipher=Twofish
[10:06] <toad_> RequestInterval=600000.0
[10:06] <toad_> DataLength=10052c
[10:06] <toad_> UniqueID=7a81b60ae64dcca7
[10:06] <toad_> Data
[10:06] <toad_> aha
[10:06] * toad_ adds them up properly
[10:06] <toad_> 316 bytes for DataReply, including Storables
[10:07] <toad_> 144 bytes for DataRequest
[10:07] <sanity> what about the DataSource? and its initial estimator?
[10:07] <toad_> 81 bytes for Accepted
[10:07] <toad_> sanity: yes, StoreData can be huge
[10:11] <toad_> phew
[10:12] <toad_> f*ck
[10:12] <toad_> FORTY TWO KILOBYTES ?!?!?!
[10:12] <toad_> woah
[10:12] <sanity> for what?
[10:14] * Hory (~Miranda@81.196.25.110) Quit (Read error: 104 (Connection reset by peer))
[10:15] <toad_> hmmm
[10:15] <toad_> 34kB for another one
[10:16] <toad_> can be gzipped down to 5kB
[10:16] <toad_> a binary format would probably compress even further
[10:16] <toad_> also I'm not convinced that the current StoreData mechanism is optimal
[10:16] <toad_> I don't know what to replace it with though
[10:17] <toad_> the above numbers are for StoreData
[10:17] <toad_> well the 34kB is for estimators only
[10:18] <toad_> the rest of the StoreData adds up to ~ 1kB
[10:18] * Gunnar (~gunnar@217-13-11-42.dd.nextgentel.com) Quit (Remote closed the connection)
[10:18] <toad_> so the total overhead is about 7kB
[10:18] <toad_> if we compress the StoreData :)
[10:19] <toad_> (using gzip)
[10:20] <toad_> one other point: if we had fixed file size, we'd have simpler estimators
[10:20] <toad_> therefore we can cut it a bit further
[10:21] <toad_> if we can get the overhead down to 4kB, then I think 16kB filesize wouldn't be too unreasonable
[10:21] <TheSeeker> 7k overhead on a 4k file seems like a bad plan :P
[10:22] <toad_> TheSeeker: :)
[10:24] <hirvox> btw, what happened to thelema's binary presentation layer?
[10:24] <toad_> hirvox: my point exactly
[10:24] <toad_> right now it's not really needed
[10:25] <toad_> however if we cut the max file size to 256kB, it'd be significantly more useful
[10:25] <toad_> if we fixed the file size at 32kB, it'd be vital
[10:26] <hirvox> couldn't we just gzip everything?
[10:26] <toad_> yup, that would achieve a reasonable gain
[10:26] <toad_> but esp. for small messages, it'd be better to have binary presentation
[10:27] <KenMan> doesn't having more pieces mean greater odds of not finding them all ?
[10:27] <toad_> KenMan: hrrm?
[10:27] <toad_> splitfiles have been redundant for aaages
[10:27] <hirvox> Kenman:yes, that's why it would need transparent FEC
[10:27] * toad_ is posting a summary of his arguments
[10:28] <KenMan> whaddaya mean re: splitfiles ?
[10:37] <toad_> KenMan: splitfiles are redundant
[10:38] <toad_> they contain redundancy
[10:38] <toad_> you need 67% of the blocks inserted to get the file
[10:39] * iip_i2p (~changate@host.teitel.net) Quit (Remote closed the connection)
[10:40] <KenMan> but at the cost of inserting 133% by volume ?
[10:40] <toad_> 150%
[10:40] <toad_> 1.5 * 0.67 = 1.0
[10:41] <KenMan> just to play devil's advocate, how does this differ from inserting the file 153% more (ie a higher HTL) ?
[10:42] <KenMan> okay, FEC allows you fetch almost any 67% of the blocks... i think it is a win. But I'm unsure...
[10:44] <KenMan> so many issues to consider, so little time in which to ponder the meaning of life (and FreeNet) ...
[10:44] <toad_> KenMan: well, if we have 150 blocks
[10:44] <TheSeeker> I thought FEC allowed *any* combination of the correct number of blocks...
[10:44] <toad_> what is the probability of success needed for all of them to be available?
[10:45] <toad_> if the overall psuccess must be 0.95
[10:45] <KenMan> it depends on the amount of redundancy on the network, i suppose
[10:45] <toad_> then each one must have a psuccess of at least 99.97%
[10:46] <KenMan> not if you think outside the cube - if file A has 200 copies floating around, you only need 1/200 success (to redefine success)...
[10:46] <KenMan> anyway, i stop distracting you from the good stuff.
[10:46] * Hory (~Miranda@81.196.25.110) has joined #FreeNet
[10:47] <toad_> it is unrealistic to expect that level of success rate
[10:47] <toad_> in something like freenet we're bound to have SOME failures
[10:47] <toad_> even in a near ideal situation
[10:48] <KenMan> time for me to start up another pre-spec'ed testnet ? Toad- you want my barcode maker program(s) ??
[10:49] <toad_> hmmm
[10:49] <toad_> speaking of which...
[10:49] <toad_> 7/21/04 9:00:00 PM BST 44 42 0.9545454545454546
[10:49] <toad_> 7/21/04 10:00:00 PM BST 627 543 0.8660287081339713
[10:49] <toad_> 7/21/04 11:00:00 PM BST 581 507 0.8726333907056799
[10:49] <toad_> 7/22/04 12:00:00 AM BST 532 458 0.8609022556390977
[10:49] <toad_> 7/22/04 1:00:00 AM BST 427 367 0.8594847775175644
[10:49] <toad_> 7/22/04 2:00:00 AM BST 405 347 0.8567901234567902
[10:49] <toad_> 7/22/04 11:00:00 AM BST 94 80 0.851063829787234
[10:49] <toad_> 7/22/04 12:00:00 PM BST 58 47 0.8103448275862069
[10:49] <toad_> 7/22/04 1:00:00 PM BST 32 26 0.8125
[10:49] <toad_> 7/22/04 2:00:00 PM BST 37 31 0.8378378378378378
[10:49] <toad_> 7/22/04 3:00:00 PM BST 36 28 0.7777777777777778
[10:49] <toad_> 7/22/04 4:00:00 PM BST 32 25 0.78125
[10:50] <KenMan> why the volume so durned screwy ???
[10:51] <toad_> this is routingSuccessRatio
[10:51] <toad_> that's why
[10:51] * iip_i2p (~changate@host.teitel.net) has joined #freenet
[10:51] <KenMan> oh, WOW.
[10:51] <KenMan> wait, on a provider node or the requestor node ?
[10:51] <toad_> the requestor node
[10:52] <KenMan> so why does volume drop off like that ? you doing pushes and pulls or something ?
[10:53] <KenMan> barcode support has been removed from fred for a long time now... :(
[10:54] <KenMan> hey , are you sitting at 3pm or 4pm now ? I would guess 4, but maybe British summer time has flipped already. I doubt it though.
[10:54] <toad_> 4pm
[10:54] <toad_> pulling data
[10:54] <toad_> volume drops off because most of it is cached
[10:55] <toad_> i'm more worried about the ratio
[10:55] <toad_> it apparently drops off
[10:55] * TheSeeker (BluJaguar@h-69-3-210-214.lsanca54.dynamic.covad.net) Quit (Read error: 60 (Operation timed out))
[10:55] <toad_> and it never gets REALLY high for long
[10:55] <KenMan> are you pulling a decreasing volume on purpose ?
[10:55] <toad_> no, stuff is cached
[10:56] <KenMan> well, look at it this way - you made it twice as good as it was, and if it doesn't make much difference on the big nets, then we still have a long way to go.
[10:56] <KenMan> shrink that DS, or delete it every hour, or something !!!
[10:56] <KenMan> just try to keep the volume consistent
[10:57] <toad_> well, the upside is, the requestSuccessRatio is really high
[10:57] <KenMan> meaningless, as you already said.
[10:57] <toad_> steadily rising and now above 99%
[10:57] <toad_> yeah
[10:57] <toad_> i need to cut the store size to 101M
[10:58] <toad_> ah
[10:58] <toad_> the total data is only 200MB
[10:58] <KenMan> maybe grow the keypool or something. Yeah, that would do it !!!
[10:58] <KenMan> what size keys are you using ?
[10:59] <toad_> sure but that would take fscking ages
[10:59] <toad_> 32k iirc
[10:59] <toad_> or was it 64?
[10:59] <toad_> not the 256 the scripts did originally
[10:59] <toad_> i reduced for some reason
[10:59] <toad_> ah yes to speed it up
[10:59] <KenMan> just destroy the requestor's store after every 20 requests (that won't take very long)
[11:00] <toad_> nah, if i set the storesize to 101MB, and the total is 197MB, then it won't get good from the store
[11:00] <KenMan> sounds like your effort to speed it up has slowed it down :(
[11:00] <toad_> inbound requests graphs were quite impressive
[11:01] <KenMan> you mean well spec'ed ?
[11:01] <toad_> yeah
[11:01] <KenMan> or volume of queries at the point of origination ?
[11:01] <toad_> hmm?
[11:01] <KenMan> n/m
[11:02] <KenMan> go fix up your testnet :) bb4hr
[11:07] <toad_> i've fixed it
[11:08] <toad_> i'm merging
[11:12] <toad_> argh
[11:12] <toad_> bbiab
[11:12] * toad_ (toad@82-32-16-91.cable.ubr03.azte.blueyonder.co.uk) Quit ("using sirc version 2.211+KSIRC/1.3.10")
[11:14] * toad_ (toad@82-32-16-91.cable.ubr03.azte.blueyonder.co.uk) has joined #freenet
[11:15] * someone (chili@c211-30-206-56.rivrw3.nsw.optusnet.com.au) has joined #freenet
[11:15] * someone (chili@c211-30-206-56.rivrw3.nsw.optusnet.com.au) has left #freenet
[11:20] * ShaunMacPherson (~ShaunMac@Sudbury-HSE-ppp3976660.sympatico.ca) has joined #freenet
[11:22] * toad_ starting up pre-5085...
[11:23] <toad_> That Which w
[11:23] <toad_> May Be 5085
[11:26] <toad_> and of course the first thing it does it ditch all my old estimators
[11:26] <toad_> even though i'm not merging the routing changes...
[11:26] <toad_> there were some minor ones
[11:55] * Dianka (~Diana@nskfw1.beelinegprs.com) has joined #freenet
[11:55] <Dianka> hallo!!!!
[11:55] <toad_> hi
[11:56] <Dianka> sprechen sie deutsch?
[11:57] <toad_> nein
[11:58] <toad_> 40/43 so far on the localtestnet... since 5:00
[11:59] <toad_> which is ~ 93%
[12:00] <toad_> Iakin: here?
[12:03] * ixitxachitl_ (dn32418@d66-183-170-199.bchsia.telus.net) Quit (Remote closed the connection)
[12:25] * Dianka (~Diana@nskfw1.beelinegprs.com) Quit (Connection timed out)
[12:27] <toad_> somebody once said that freenet goes in cycles of gradually getting better and then a release which breaks everything
[12:27] <toad_> and that the release that breaks everything was long overdue
[12:27] * toad_ is about to satisfy the prophecy... ;)
[12:27] * jayo (jcl@ool-18bf6dac.dyn.optonline.net) has joined #freenet
[12:28] * jayo is now known as jay
[12:29] <toad_> hi jay
[12:29] <jay> hey man
[12:29] <toad_> <toad_> somebody once said that freenet goes in cycles of gradually getting better and then a release which breaks everything
[12:29] <toad_> <toad_> and that the release that breaks everything was long overdue
[12:29] <toad_> toad_ is about to satisfy the prophecy... ;)
[12:29] <toad_> s/satisfy/fulfill
[12:30] <toad_> by releasing 5085
[12:30] <toad_> very soon
[12:30] <sanity> fun fun fun!
[12:32] <toad_> yup, it's definitely faster
[12:32] <toad_> the new code draws the graphs faster
[12:32] <toad_> that's why it felt slow last time
[12:37] <jay> i think it's slow cause it's written in java
[12:37] <jay> everyone knows assembler is the best language ever
[12:37] <jay> eheh
[12:37] * jay trolls a little
[12:38] <toad_> :)
[12:38] <jay> toad_: i have a Q on inserts..
[12:38] <jay> i just inserted a cd of mp3's at htl=0
[12:39] <jay> when i re-insert @ htl=3, it moves rather fast..
[12:39] <jay> is it actually sending the insert requests outside my node?
[12:39] <greycat> only if pcaching ate some of the mp3s
[12:39] <jay> or is it checking the local datastore and stopping after it finds it already inserted there?
[12:39] <greycat> if they all "stuck" to your node, then no, they collide at localhost and You Lose.
[12:40] <jay> hrm
[12:40] <jay> is that what RemoveLocalKey is for then?
[12:40] <greycat> if you're feeling particularly energetic, you can badger toad to fix the "skip local data store" and/or "remove key from local data store" options on FEC splitfiles
[12:40] <jay> does skip local work for clientput also?
[12:40] <jay> i thought skiplocal was for ClientGet only
[12:41] <greycat> whichever version of it I'm using in FT seems to work fine for single files. But when you check the magic box on the FEC splitfiles in the web gateway, it Fails. Every time.
[12:41] <greycat> (the magic box is on retrieval)
[12:41] <jay> fcpget isn't working atm but fcpput is almost perfect
[12:42] <toad_> greycat: umm, i did
[12:42] <greycat> toad_: not in stable yet...
[12:42] <jay> greycat: do you mean that skipLocal and RemoveLocalKey apply to both ClientPut and ClientGet ?
[12:42] <toad_> on ClientPut, it's fixed
[12:42] <toad_> at least
[12:42] <toad_> not sure about ClientGet, didn't know there was a problem there
[12:42] <toad_> and yes, I'm busy merging it :)
[12:42] <greycat> ftfcp.c: if (0 != sk_printf(&f->conn, "RemoveLocalKey=true\n")) {
[12:42] <greycat> jay: that one.
[12:42] <jay> i can't speak about ClientGet really.. ive been focused on inserts
[12:43] <jay> greycat: that's ClientPut yes?
[12:43] <greycat> yes
[12:43] <jay> ok
[12:43] <jay> so ClientPut cannot SkipLocal, only RemoveLocal ?
[12:43] <greycat> I never tried SkipLocal
[12:44] <jay> it seems that SkipLocal would be for fcpget
[12:44] <jay> err fcpget = ClientGet
[12:44] <jay> now im re-inserting the CD with RemoveLocalKey=true.. @ htl=3...
[12:45] <greycat> all I know for sure is that in build 5084, (1) FT can't do FEC inserts at all, (2) the web gateway can't skipDS on FEC gets (blows up horribly), (3) freenet.client.cli.Main can't either.
[12:45] <jay> greycat: fcpput is inserting FEC files rather well
[12:46] <toad_> RemoveLocal == SkipLocal
[12:46] <jay> it seems to be working now with RemmoveLocalKey=true, however a little slower
[12:46] <toad_> it's implemented as skip local
[12:46] <toad_> not remove local
[12:46] <jay> toad_: ah
[12:46] <greycat> don't worry, keys get eaten up by pcaching just fine...
[12:46] <jay> so it never deletes just skips
[12:46] <jay> well it always attempts to Skip, and not actually delete the local key
[12:47] <jay> toad_: should fcptools then treat them both as -S --skip-local for Put and Get ?
[12:47] <greycat> or maybe pcaching only deletes the keys I don't want deleted. It could be sentient for all I know....
[12:47] <jay> greycat: eheh
[12:48] <jay> i don't understand pcaching at all
[12:48] <jay> there's much about the world i don't understand ;)
[12:48] <greycat> I understand the basic reason behind it, but I never looked at the algorithm he's using to see if it's even sane, let alone implemented correctly
[12:48] <jay> i just got a failed insert after 5 exhausted retries.. restarting with retry count=15
[12:49] <jay> toad_: this time im getting RouteNotFound, other than timeouts on Pending messages, so perhaps this is much better
[12:50] <jay> err i just noticed fcpput isn't cleaning all the tempfiles.. ok there's 1 missing line of code somewhere..
[12:50] * Lemoni (~Lemoni@CPE-203-45-11-225.vic.bigpond.net.au) has joined #freenet
[12:51] <toad_> <greycat> or maybe pcaching only deletes the keys I don't want deleted. It could be sentient for all I know....
[12:51] <toad_> :)
[12:52] <toad_> <greycat> I understand the basic reason behind it, but I never looked at the algorithm he's using to see if it's even sane, let alone implemented correctly - so go look at it :)
[12:52] <toad_> src/freenet/node/Node.java
[12:52] <toad_> method called shouldCache iirc
[12:52] <greycat> you're not helping me get my actual Work done :-/
[12:52] <toad_> :)
[12:53] * toad_ is doing some test inserts prior to committing 5085
[12:53] <jay> im getting lots of RNF's and Restarted msgs
[12:54] <jay> but no socket timeouts.. so perhaps my node isn't working at full strength
[12:54] <toad_> can anyone get CHK@zzutSBHCm0Im3yxSZcLbqFyOO0oNAwI,2w4V2mUvPy2VhqCIhGanDQ/start-freenet.sh on stable?
[12:54] * ixitxachitl_ (dn32418@d66-183-170-199.bchsia.telus.net) has joined #freenet
[12:55] * toad_ inserting another one... and an mp3
[12:55] <jay> CHK@SSR2G1~sEkCcvJsKpsYSlud3KRkLAwI,uwMOfqtTax3GhnLefHIDgA/04-Slip.mp3 (unstable)
[12:56] <toad_> sorry, my unstable node is down atm
[12:57] <toad_> greycat: got stable? please fetch CHK@zzutSBHCm0Im3yxSZcLbqFyOO0oNAwI,2w4V2mUvPy2VhqCIhGanDQ/start-freenet.sh
[12:57] <greycat> toad_: DNF at 15 on the first attempt by my node
[12:57] <toad_> thanks
[12:57] <toad_> please retry
[12:57] * boardumb (~someone@c211-28-175-53.mckinn1.vic.optusnet.com.au) has joined #freenet
[12:57] <greycat> it's in a web browser so it's retrying
[12:58] <toad_> well tell me if it gets it okay?
[12:58] <greycat> DNF@19...
[12:58] <toad_> not too promising...
[12:59] <toad_> but then this is what i'm trying to fix - all the stuff fixed in 5085 is causing a network of 5084s to suck
[12:59] * BrianR (brianr@2002:1893:ac33:0:0:0:0:1) has joined #freenet
[12:59] <toad_> at least, that's the optimistic perspective
[12:59] <toad_> the pessimistic perspective is that 5085 will shock the network for weeks and after that some hideous problem we don't know about will ensure stable continues not to work
[13:02] <toad_> hehe, old CofE post:
[13:02] <toad_> I long for the day when all I have to bitch about as far as Freenet is concerned is the HTML filters stifling user creativity. Could that day be soon ? I certainly bloody hope so.
[13:03] <toad_> 5th April 2003
[13:03] <toad_> :)
[13:04] <toad_> hmmm
[13:05] <toad_> another insert bug :|
[13:06] <jay> im not sure FEC has anything to do with the insert problems
[13:06] <toad_> no, it probably doesn't
[13:06] <jay> in the end fcpput is calling PutBlock to insert a CHK
[13:06] <greycat> no sign of your CHK here, toad
[13:07] <jay> it's calling PutBlock 20 times.. but it's the same put routine for single files as well
[13:07] <toad_> anyway I DID fix several other insert bugs
[13:08] <jay> i believe you really
[13:11] <jay> i just received almost 20 socket timeouts on inserting a data block
[13:11] <jay> the others were RNF's
[13:11] <toad_> f*ck
[13:12] <toad_> that's on unstable?
[13:12] <toad_> CHK@tn8gGGuabP7PVHuhFoJcaN5ddW0PAwI,0fopDrpADpkvxjZ-tYyi1Q/something.java
[13:12] <toad_> somebody please fetch
[13:12] <toad_> INSERTING_BLOCKS (4/34):[1/1] queued: 11 running: 19
[13:12] <toad_> (not that one, a different one)
[13:12] <toad_> INSERTING_BLOCKS (5/34):[1/1] queued: 10 running: 19
[13:13] <toad_> the big insert hasn't had any successes yet
[13:13] <toad_> INSERTING_BLOCKS (6/34):[1/1] queued: 9 running: 19
[13:13] * toad_ is inserting a music MP3 that nobody will want and a political OGG that nobody will want either
[13:14] <toad_> (because anyone wwho cares already has it)
[13:15] <toad_> brb
[13:19] * Sugadude (foobar@64.147.163.239) has joined #freenet
[13:23] <toad_> CHK@SP~rPTElv9h~T3Dl2R5MVZvGhAkNAwI,zNg~xOy~xXEMJxmNnb6U4A/somethingelse.java
[13:23] <toad_> on stable
[13:25] * toad_ committing 5085...
[13:28] * boardumb (~someone@c211-28-175-53.mckinn1.vic.optusnet.com.au) Quit ()
[13:28] <jay> bbl
[13:32] <toad_> committed!
[13:32] * toad_ updating snapshots
[13:32] <sanity> toad: do you want to update the website or shall I?
[13:33] <toad_> sanity: I have a message that I could send you that summarizes the changes in more or less readable english (not the commit message, slightly better ;) )
[13:33] <sanity> toad: cool
[13:33] <toad_> which I'm going to send out when dodo finishes updating the snaps
[13:33] <toad_> you can base it on that if you like
[13:34] <toad_> probably not appropriate to send out a Second Coming of Freenet message on announce yet :)
[13:37] <toad_> done
[13:37] <toad_> snapshots updated
[13:38] * toad_ changes topic to 'Upgrade to 5085 | Unstable: Upgrade to 60167 (60160: network level IP detection should work now); network reset at 60151 so 60151 is mandatory | Unstable nodes must be updated daily | Stable nodes should generally be updated weekly | #freenet-chat and #freenet-politics are open for business! Off topic discussions may be directed there. | Logs: http://newton.matcmp.ncc.edu/~lockej/freenet/chanlog/'
[13:38] <toad_> sanity: sent
[13:38] <toad_> sanity: suggest you turn that into a website update
[13:46] * salahx (salahx@sc1-24.217.174.147.charter-stl.com) has joined #freenet
[13:47] <toad_> bbiab
[13:49] * Lemoni (~Lemoni@CPE-203-45-11-225.vic.bigpond.net.au) Quit (Read error: 110 (Connection timed out))
[14:12] <salahx> ugh
[14:35] <salahx> I jstu ca'nt get this NAtiveBigINterger thing to work!
[14:39] <salahx> oh I think i see why
[14:40] <toad_> a freesite for reactos is doubly surreal
[14:41] <salahx> there there one on there
[14:52] * kiloh^ (~kiloh_@ddesk-rt1.xs4all.nl) has joined #freenet
[14:52] <kiloh^> hi all
[14:52] <salahx> Hi kiloh^
[14:54] <kiloh^> anyone willing to explain how I can connect to freenet using mirc? did it before just dont remember. was smth like server 127.0.0.1 but what port?
[14:54] <salahx> you don;t conenct ot Freenet using mIRC
[14:54] <salahx> YOu;re think IIP
[14:54] <kiloh^> omg got it mixed up
[14:54] <kiloh^> thanx..!
[14:54] <toad_> :)
[14:57] * kiloh^ (~kiloh_@ddesk-rt1.xs4all.nl) Quit ()
[15:05] * Filipu (~upg269@host81-153-29-120.range81-153.btcentralplus.com) has joined #freenet
[15:05] * Toast (~k@82-34-170-133.cable.ubr03.maid.blueyonder.co.uk) has joined #freenet
[15:07] * Filipu (~upg269@host81-153-29-120.range81-153.btcentralplus.com) has left #freenet
[15:26] * interrupt (~chatzilla@64.122.23.213) has joined #freenet
[15:29] <salahx> Ther, its FINALLY working!
[15:34] <toad_> salahx: what is?
[15:34] <salahx> NAitveBigINterger support
[15:35] <toad_> ahhh
[15:36] <salahx> I compiledi t specially for Gentoo so it woudl sue the system librareis that are already optinzed and installed
[15:43] <toad_> are they?
[15:43] <toad_> gentoo automatically built the right ones?
[15:43] <salahx> they shoudl be
[15:44] * jay (jcl@ool-18bf6dac.dyn.optonline.net) Quit (".")
[15:46] <salahx> emerge gmp
[15:47] <salahx> Seems to be working
[15:49] * robilad is now known as robilad[afk]
[15:50] <salahx> I had to recompile libjbigi.so to use the shared libraries instead of the static ones
[15:54] <salahx> and then put it in the freenet-ext.jar as net/i2p/util/jbigi-linux-none.so
[16:04] * greycat (~wooledg@192.35.79.70) Quit ("Lick Bush in '04")
[16:09] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) Quit (Read error: 104 (Connection reset by peer))
[16:10] <vsalento> salahx: would you like to run the jbigi tests and tell how much faster it is compared to java implementation?
[16:16] <salahx> sure!
[16:16] * Zorix (Brandon@fl-65-40-46-197.dyn.sprint-hsd.net) has joined #freenet
[16:16] * TheSeeker (BluJaguar@h-69-3-213-194.lsanca54.dynamic.covad.net) has joined #Freenet
[16:16] <vsalento> do you know howto?
[16:17] <salahx> there we instreciton in the NAtiveBigInterger library
[16:17] <TheSeeker> were instructions?
[16:17] <salahx> yeah
[16:17] <vsalento> alright... tell the results when it is finished
[16:17] <Zorix> sheesh not much updates to stable build in a while
[16:19] <salahx> there was jsut a new verison of stable relased today!
[16:19] <vsalento> for me the native version of modpow took only 17% time compared to java impl.
[16:21] <salahx> acurtallyi tl ooks lie its harder than tis sounds, its too gleus to everything else...
[16:22] <vsalento> mean what? the tests can be run very easily... do you need help with them?
[16:22] <salahx> how do I run them then ?
[16:22] <vsalento> java -cp lib/freenet.jar:lib/freenet-ext.jar net.i2p.util.NativeBigInteger
[16:23] <vsalento> run that on freenet directory
[16:23] <salahx> ok thaT IS easy :)
[16:24] <salahx> its running now...
[16:25] <salahx> INFO: run time: 3702ms (36ms each)
[16:25] <salahx> INFO: 100 runs complete without any errors
[16:25] <salahx> native run time: 3702ms (36ms each)
[16:25] <salahx> java run time: 12919ms (127ms each)
[16:25] <salahx> native = 28.655468689527055% of pure java time
[16:25] <vsalento> is that for modpow? for my machine it takes 649ms each :(
[16:26] <salahx> yeah
[16:27] <vsalento> INFO: run time: 65627ms (649ms each)
[16:27] <vsalento> INFO: 100 runs complete without any errors
[16:27] <vsalento> native run time: 65627ms (649ms each)
[16:27] <vsalento> java run time: 379010ms (3752ms each)
[16:27] <vsalento> native = 17.31537426453128% of pure java time
[16:27] <vsalento> when comparing these values it seems that you have better CPU than my machine has :)
[16:27] <salahx> ptobably
[16:29] <Zorix> new stable today
[16:29] <Zorix> ?
[16:29] <salahx> Yep, 5085
[16:29] <salahx> Coem and get it!
[16:30] <salahx> It cram-packed full of new bugs^H^H^H^Hfeatures!
[16:33] * TheSeeker changes his listenport, kills his RT and seednodes, and grabs latest stable.
[16:34] <TheSeeker> odd, Freenet doesn't generate a seednodes.ref file if you don't have one...
[16:34] <Zorix> lol
[16:34] <TheSeeker> I suppose it just uses the RT files then if those exist and doesn't bother with the ref?
[16:43] <TheSeeker> of course, it helps to have a ref when starting a new node ?.?
[16:45] <salahx> right
[16:45] <salahx> The seednodes are only used when you hjave no valid refereces
[16:46] <TheSeeker> someone bought $13.5 million worth of AMD after hours o,o 1,075,000 shares *whistles*
[16:48] <salahx> wow
[16:48] <TheSeeker> I wish I had that kinda money :-p
[16:49] <interrupt> Thats an awful lot of eggs in one basket
[16:50] <salahx> depepdnson tehj size of the basket :)
[16:50] <salahx> Id BIll Gate bought it; that;s a drop in the bucket for him
[16:52] <Toast> whoever it was is cashing in on the release of Doom 3 in august
[16:52] <TheSeeker> I'm betting it was a fund manager covering their short.
[16:55] <salahx> ?
[16:57] * Sugadude (foobar@64.147.163.239) Quit ()
[16:59] <TheSeeker> the short interest on AMD's stock is something like 16% of the float, if the stock goes down, short make money by buying the shares they don't own back at a lower price that they borrowed them (sold) at ... the stock is likely to rebound, so smart shorts would cover (buy) while they still have a chanve to profit.
[16:59] <salahx> ahhh ok
[16:59] <TheSeeker> otherwise, they could be forced to buy back the shares at a higher price than they borrowed thm at and lose money
[17:00] <salahx> got it
[17:00] <salahx> Well, at elast you dont; ahve otworry abotu thse guys stock rebounding: http://finance.yahoo.com/q/bc?s=SCOX
[17:00] <TheSeeker> I don't use margin or try to go short/long on stocks 'cause gambling with money I dont' have just seems to dangerous to me :P
[17:00] <salahx> yeah
[17:01] <TheSeeker> I'm more intersted in http://finance.yahoo.com/q/bc?s=^soxx
[17:01] <TheSeeker> been taking a beating lately.
[17:03] <salahx> never heard of THAT index before
[17:03] <TheSeeker> semi index
[17:04] <TheSeeker> contains stuff like AMD, INTC, MOT, AMAT, TXN etc...
[17:04] <salahx> i didfuer that' waht it woudl probabyl contain...
[17:06] * TheSeeker decodes that successfully
[17:08] * KenMan wisely chooses not to try :)
[17:09] * greycat (rfc1413@wooledge.org) has joined #freenet
[17:12] <vsalento> odd... i have inbound request logging enabled and there is one node from where I have accepted 167 requests although it has only sent 165.
[17:13] * guido^pe (~unknown@dsl-082-082-158-159.arcor-ip.net) Quit (Remote closed the connection)
[17:13] <vsalento> for other nodes the sent request and accepts are same.
[17:13] <salahx> I ahve NPTL enable,d how can I found out if the same thing is hapopenign to me that;s also happinign to Toad ?
[17:17] * Toast (~k@82-34-170-133.cable.ubr03.maid.blueyonder.co.uk) Quit ()
[17:27] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) has joined #freenet
[17:28] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) Quit (Read error: 54 (Connection reset by peer))
[17:32] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) has joined #freenet
[17:34] * Filipu (Filipu@host81-153-31-185.range81-153.btcentralplus.com) has joined #freenet
[17:38] <TheSeeker> yay. looks like people are recognizing my datastore specialization at C ...
[17:38] <TheSeeker> (stable)
[17:39] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) Quit ("using sirc version 2.211+KSIRC/1.3.10")
[17:39] <oierw> how did you specialize?
[17:39] * interrupt (~chatzilla@64.122.23.213) Quit ("Chatzilla 0.9.64b [Mozilla rv:1.7/20040707]")
[17:39] <TheSeeker> deleted all the keys that weren't 1-c* in my datastore...
[17:41] <oierw> so nothing crazy like deleting most of the entries that didn't also have a specialization around a-e then
[17:42] * ysothoth (~ysothoth@68-174-22-46.si.rr.com) has joined #freenet
[17:44] <TheSeeker> nah, I'm too lazy to shape a curve...
[17:45] <TheSeeker> I only know about one other node that's latest stable >.o
[17:45] <greycat> Fred,0.5,STABLE-1.50,5082 44
[17:45] <greycat> Fred,0.5,STABLE-1.50,5084 311
[17:45] <greycat> Fred,0.5,STABLE-1.50,5085 18
[17:46] <TheSeeker> yay, I found yesterday's FIND
[17:46] <TheSeeker> I just started a new node, so I'm using the default seednodes.ref... only know 107 nodes atm...
[17:46] <greycat> I'd like to find tomorrow's Greg_Wooledge. It *says* it was inserted.... :-/
[17:47] <greycat> ah well, perhaps I could do a --delete reinsert since I just upped to '85.
[17:47] <iip_i2p> <Sonax> Toad (or, anyone): Should the new stable build not be anounced to the support list (or have i gone blind?)
[17:47] <TheSeeker> it's announced on devl :P
[17:47] <mazzanet> and announce :P
[17:47] <greycat> don't most of the devl readers run unstable?
[17:47] <oierw> but there isn't a website post either
[17:48] <mazzanet> there no ned to
[17:48] <TheSeeker> people runnign nodes should see a bold new version on their fproxy start page
[17:48] <TheSeeker> an announcement would still be nice.
[17:49] <TheSeeker> SSK@wlUk5niRuNDUL4lzvdHu6DamhW8PAgM/Greg_Wooledge// ?
[17:49] <iip_i2p> <Sonax> People should read devl, but that's not going to happe either :) Spread the word with all means at your disposal...
[17:49] <greycat> That's today's. Tomorrow's would be SSK@wlUk5niRuNDUL4lzvdHu6DamhW8PAgM/Greg_Wooledge//?date=20040723
[17:49] <greycat> or the perl-printf-%x version
[17:50] <mazzanet> SSK@wlUk5niRuNDUL4lzvdHu6DamhW8PAgM/Greg_Wooledge//?date=20210523
[17:50] <mazzanet> :P
[17:50] <greycat> I haven't inserted into the year 2021 yet.
[17:51] <mazzanet> tomorrow, everyone should insert a year 2021 version :P
[17:51] <salahx> you need to ehbcasue that' howl ong it tkaes to downlaod something on Freenet :)
[17:52] * greycat does a t -d --delete -l 10 -i 40 putsite KSK@$$.$RANDOM.$RANDOM Greg_Wooledge/ .... again. And hopes toad really did fix insertion in '85.
[17:52] <greycat> err, "ft -d ..." not "t -d ...". Lost the first letter.
[17:53] <greycat> Lots and lots of "SUCCESS" and not so many "COLLISION". That's bad.
[18:08] * Herkamire (~jason@h000094d30ba2.ne.client2.attbi.com) has joined #freenet
[18:08] <Herkamire> can I get freenet not to use more than 100MB of RAM?
[18:11] <Herkamire> I'm sick of constant swapping
[18:13] <greycat> it's a memory pig
[18:15] <Herkamire> I'd like to shutdown freenet for a bit, then start it up later today.
[18:16] <Herkamire> is there something I could/should do to make it not forget all the lovely nodes it's found over the last 24 hours?
[18:22] <Herkamire> I downloaded my first big file from freenet today. 5.5MB. downloaded I think 128 parts
[18:23] <TheSeeker> 5.5 MB shouldn't be 128 parts o.O
[18:23] <mikeDOTd> Herkamire: shutting down your node won't make it lose it's routing table
[18:23] <Herkamire> mikeDOTd: awesome
[18:27] <TheSeeker> hmm, I'm getting a large amount of requests for b, even though I'm best at giving c keys...
[18:41] * nextime (~nextime@danex.i-m-c.it) Quit ("nextime has no reason")
[18:49] <greycat> TheSeeker: if you used to have lots of b keys, then other nodes will think you specialize in b, not c
[18:49] <greycat> this should adjust over time, of course
[18:57] <TheSeeker> well, I'll take a nap and check back in a while...
[19:19] * Zorix- (Brandon@fl-65-40-46-197.dyn.sprint-hsd.net) has joined #freenet
[19:23] <Ash-Fox> Oh hell, this is just insane: http://slashdot.org/article.pl?sid=04/07/22/2156240
[19:24] <greycat> My country's political system is corrupted. And I can't find the psck command.
[19:26] * nextime (~nextime@danex.i-m-c.it) has joined #freenet
[19:28] <greycat> toad_: How long is an FCP insert supposed to sit without result before I can conclude that inserts still hang in 5085?
[19:28] * Zorix (Brandon@fl-65-40-46-197.dyn.sprint-hsd.net) Quit (Read error: 60 (Operation timed out))
[19:29] <greycat> I've got one ftput process (single file insert over FCP) that was started at 18:06 and it's now 19:45.
[19:31] <greycat> I killed it and it started a new attempt for the same file. That one succeeded (SUCCESS not COLLISION) about 1 minute later.
[19:32] <greycat> When I killed the other one (which had been running since 18:11), its successor finished with COLLISION a few seconds later.
[19:32] <toad_> greycat: ugh :<
[19:33] <toad_> what are you using to insert?
[19:33] * Hory (~Miranda@81.196.25.110) Quit ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
[19:33] <greycat> uh, FT version 0.2.4-and-a-half if you want to call it that.
[19:34] <greycat> I want to make some more changes to it before I release it as 0.2.5. The most important of those is to add a timeout so that when an ftput process has been running for, say, 10 minutes it will self-destruct instead of lingering like this.
[19:38] <toad_> well i need to see it linger
[19:38] <toad_> all the other clients have timeouts :)
[19:38] <toad_> does it get no message whatsoever in that time?
[19:39] <toad_> it's possible it would get Success without getting Pending?
[19:39] <greycat> I wasn't running with full debugging on. It's *very* noisy that way...
[19:39] <toad_> well i'll have a look soon
[19:39] <toad_> bbl zzz
[19:57] -NickServ- This nickname is owned by someone else
[19:57] -NickServ- If this is your nickname, type /msg NickServ IDENTIFY <password>
[20:00] -lilo- [Global Notice] Hi all. Services have been restored. Please be on the lookout for any problems with nickserv and chanserv data. Feel free to message if you see anything unusual. Thanks!
[20:08] * Zorix- (Brandon@fl-65-40-46-197.dyn.sprint-hsd.net) Quit (Read error: 60 (Operation timed out))
[20:21] <iip_i2p> <_oOo_> I don't have any inbound connections with my 5085 whereas it got some before the update (done minutes after the SF update). Is this normal ? Am I the only one observing this ?
[20:21] * greycat (rfc1413@wooledge.org) Quit ("Lick Bush in '04")
[20:33] <iip_i2p> <_oOo_> Ok, I take it back ^^ Found I was somewhat guilty...
[20:37] * Gunnar (~gunnar@34.80-203-209.nextgentel.com) has joined #freenet
[20:51] * piranha (~piranha@192.220.223.23) has joined #freenet
[21:06] <TheSeeker> firewalled much?
[21:14] <iip_i2p> <_oOo_> Wrongly configured virtual DNS update software ^^
[21:39] * Zorix (Brandon@fl-65-40-46-197.dyn.sprint-hsd.net) has joined #freenet
[21:41] * FridlekhToo (BluJaguar@h-66-167-104-40.lsanca54.dynamic.covad.net) has joined #Freenet
[21:50] <FridlekhToo> umm, is that stuff open source? http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=1251&item=3689579568&rd=1&ssPageName=WDVW
[21:57] * TheSeeker (BluJaguar@h-69-3-213-194.lsanca54.dynamic.covad.net) Quit (Read error: 110 (Connection timed out))
[22:00] * FridlekhToo is now known as TheSeeker
[22:02] * Lemoni (~Lemoni@munyoo.cs.mu.OZ.AU) has joined #freenet
[22:06] * Lemoni (~Lemoni@munyoo.cs.mu.OZ.AU) Quit (Client Quit)
[22:08] <iip_i2p> <gott> Fee. Fie. Foe. Fum.
[22:10] <TheSeeker> <.< >.> ^?^ *hides*
[22:11] <iip_i2p> <gott> Hello, TheSeeker.
[22:18] <Herkamire> TheSeeker: hehe, sure is :)
[22:18] <Herkamire> valuable stuff :)
[22:18] * piranha_ (~piranha@192.220.223.10) has joined #freenet
[22:19] <TheSeeker> got several of these:
[22:19] <TheSeeker> Jul 22, 2004 7:29:49 PM (freenet.node.rt.NGRoutingTable, main): Caught java.io.IOException: Value out of range: 4500000.0 deserializing a NodeEstimator for DataObjectRoutingMemory:tcp/24.158.21.13:46696, sessions=1, presentations=3, ID=DSA(0100 0ab8 f176 3e02 7070 3833 9bd0 2060 e641 05d5), version=Fred,0.5,STABLE-1.50,5084:01000ab8f1763e02707038339bd02060e64105d5
[22:19] <TheSeeker> java.io.IOException: Value out of range: 4500000.0
[22:20] <TheSeeker> O.o and then this:
[22:20] <TheSeeker> Jul 22, 2004 7:31:40 PM (freenet.node.Main, main): Skipped bad NodeReference while reading seed nodes
[22:20] <TheSeeker> freenet.node.BadReferenceException: NodeReference self signature check failed.
[22:21] <TheSeeker> damn, it's taking forever to start up...
[22:23] * piranha (~piranha@192.220.223.23) Quit (No route to host)
[22:37] * thelema|aikido is now known as thelema
[23:13] <Ash-Fox> You know, I'm getting pretty anoyed at the people who think PHP is so great... if it is so great, why did deviantart change from php to perl... why aren't there any sites like slashdot.org written in php (with the same amount of traffic)
[23:19] <Herkamire> php sucks for more complex stuff
[23:19] <Herkamire> and perl is very fast
[23:19] <Herkamire> PHP is fun for tinkering because it has support for everything emaginable built in
[23:21] * tessier_ is having a good tiem with python
[23:21] <tessier_> Ash-Fox: Sites with as much traffic as slashdot are few.
[23:22] <Herkamire> I worked with PHP for a couple years, and I thought it was cool, but after doing mod_perl for just a couple months, I couldn't stand php anymore
[23:22] <Herkamire> I've had fun with python too.
[23:22] <Herkamire> but it's slow slow slow
[23:23] <Herkamire> I think python is about 10X slower than perl
[23:26] <Herkamire> I have not had speed problems with PHP except one program I wrote recently
[23:27] <tessier_> Why would it be 10x slower than perl
[23:27] <tessier_> ?
[23:28] <Herkamire> different design goals I guess
[23:28] <Herkamire> perl is designed to be quick and dirty
[23:28] <Herkamire> python is designed to be a clean OO language
[23:30] <tessier_> I don't see why an OO architecture would affect speed since it's all just references anyway
[23:34] <Ash-Fox> Python uses alot of CPU
[23:34] <Ash-Fox> That's the only flaw I found in Python
[23:35] <Ash-Fox> I quite like python however
[23:38] <Herkamire> I like python too, I just don't have a use for it
[23:39] <tessier_> I'm learning it to use with zope to build some complex web apps
[23:41] <Ash-Fox> zope is pretty neat, but difficult to get started with
[23:43] <tessier_> Ash-Fox: Indeed.
[23:45] * jmercier (~jmercier@4.10.164.111) has joined #freenet
[23:45] * jmercier (~jmercier@4.10.164.111) Quit (Remote closed the connection)
[23:47] <Herkamire> I'm curious about zope, but I never bothered to find out too much about it.
[23:47] <tessier_> I've been curious for years. Now I actually have a reason to learn it.
[23:47] <Herkamire> I read about some things it does, but we already do those things quite easily
[23:52] <tessier_> It does an awful lot.
[23:52] <tessier_> It's equivalent to IBM WebSphere etc.
[23:52] * thelema is now known as thelema|away
[23:52] <Herkamire> I can't think of anything I want that we don't have php doing already
These logs were automatically created by Jay Oliveri with his gimp hapi on irc.freenode.net.
